Artificial Intelligence and FLIP Panometry-Automated Classification of Esophageal Motility Patterns

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Functional lumen imaging probe (FLIP) panometry assesses esophagogastric junction and esophageal motility during endoscopy.
• Analysis of FLIP panometry data, guided by the Dallas Consensus, is complex.
• Limited research exists on artificial intelligence (AI) applications in FLIP panometry.

Purpose of the Study:

• To develop an AI model for automated classification of motility patterns in FLIP panometry exams.
• To evaluate the performance of machine learning models in identifying pathological FLIP panometry patterns.

Main Methods:

• 105 FLIP panometry exams from five international centers were analyzed.
• Machine learning models were trained and validated using a patient-split design.
• Performance was assessed using accuracy and area under the receiver-operating characteristic curve (AUC-ROC).

Main Results:

• An AdaBoost Classifier achieved 84.9% accuracy (AUC-ROC 0.92) for pathological patterns.
• Random Forest identified esophagogastric junction opening disorders with 86.7% accuracy (AUC-ROC 0.973).
• Gradient Boosting Classifier detected contractile response disorders with 86.0% accuracy (AUC-ROC 0.933).

Conclusions:

• A machine learning model accurately classified FLIP panometry exams according to the Dallas Consensus.
• AI-driven FLIP panometry offers potential to revolutionize exam standardization and accessibility.
• This technology could significantly transform patient management through optimized diagnostic accuracy.